Dataset Info¶

This dataset comprises information of 25000 mutual funds in the United states. Various attributes related to the mutual fund have been described and these attributes will be used for making decisions on the rating of the mutual fund by GreatStone which is a top mutual fund rating agency. The following files are provided in the form of CSVs. These files contain various attributes related to the mutual fund. Please find the following files for the same: bond_ratings, fund_allocations, fund_config, fund_ratios, fundspecs, other specs, return_3year, return_5year, return_10year.

Domain:¶

Mutual Fund - Finance

Objective:¶

The goal of this hackathon is to predict GreatStone’s rating of a mutual fund. In order to help investors decide on which mutual fund to pick for an investment, the task is to build a model that can predict the rating of a mutual fund. The various attributes that define a mutual fund can be used for building the model

###  Import the necessary libraries
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import scipy
import pandas_profiling
sns.set(rc={'figure.figsize':(13.7,8.27)}) # setting constant to increase seaborn graph sizes

# Read the data as a data frame
#bond_ratings consists of 12 columns which provide information on the bond rating percentage allocation of the mutual funds
#The tag column is a unique identifier and is also the same as the id.(i.e tag = id)
bond_ratings = pd.read_csv('Hackathon_Files/external/bond_ratings.csv')
pandas_profiling.ProfileReport(bond_ratings)

#fund_allocations consists of 12 columns which provide information on the sector wise percentage allocation of the mutual funds
fund_allocations = pd.read_csv('Hackathon_Files/external/fund_allocations.csv')
pandas_profiling.ProfileReport(fund_allocations)

#fund_config comprises of 4 columns which comprise the metadata of the mutual funds
fund_config = pd.read_csv('Hackathon_Files/external/fund_config.csv')
#pandas_profiling.ProfileReport(fund_config)
print(fund_config.info())
fund_config.describe().transpose()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 25000 entries, 0 to 24999
Data columns (total 4 columns):
category          25000 non-null object
parent_company    25000 non-null object
fund_id           25000 non-null object
fund_name         25000 non-null object
dtypes: object(4)
memory usage: 781.3+ KB
None

#fund_ratios consists of 8 columns which provides information on various fundamental ratios that describe the mutual funds
fund_ratios = pd.read_csv('Hackathon_Files/external/fund_ratios.csv')
pandas_profiling.ProfileReport(fund_ratios)

#fund_specs contains 9 columns which give information about the specifications of the mutual funds
fund_specs = pd.read_csv('Hackathon_Files/external/fund_specs.csv')
pandas_profiling.ProfileReport(fund_specs)

#other_specs contains 43 columns which give information of the other aspects of the mutual funds
other_specs = pd.read_csv('Hackathon_Files/external/other_specs.csv')
pandas_profiling.ProfileReport(other_specs)

#return_3years contains 17 columns which give information about 3 year return and ratios
return_3year = pd.read_csv('Hackathon_Files/external/return_3year.csv')
pandas_profiling.ProfileReport(return_3year)

#return_5years contains 17 columns which give information about 5 year return and ratios
return_5year = pd.read_csv('Hackathon_Files/external/return_5year.csv')
pandas_profiling.ProfileReport(return_5year)

#return_10years contains 17 columns which give information about 10 year return and ratios
return_10year = pd.read_csv('Hackathon_Files/external/return_10year.csv')
pandas_profiling.ProfileReport(return_10year)

Attribute Information:¶

default 1 - yes 0 - no

account_check_status: (qualitative)

     Status of existing checking account
         A11 :      ... <    0 DM  (DM - Deutsch Mark)
     A12 : 0 <= ... <  200 DM
     A13 :      ... >= 200 DM /
       salary assignments for at least 1 year
         A14 : no checking account

duration_in_month: (numerical)
```
    Duration in month
```

credit_history: (qualitative)

    Credit history
    A30 : no credits taken/
      all credits paid back duly
        A31 : all credits at this bank paid back duly
    A32 : existing credits paid back duly till now
        A33 : delay in paying off in the past
    A34 : critical account/
      other credits existing (not at this bank)

purpose: (qualitative)

    Purpose
    A40 : car (new)
    A41 : car (used)
    A42 : furniture/equipment
    A43 : radio/television
    A44 : domestic appliances
    A45 : repairs
    A46 : education
    A47 : (vacation - does not exist?)
    A48 : retraining
    A49 : business
    A410 : others

credit_amount: (numerical)
```
    Credit amount
```

savings: (qualitative)

    Savings account/bonds
    A61 :          ... <  100 DM
    A62 :   100 <= ... <  500 DM
    A63 :   500 <= ... < 1000 DM
    A64 :          .. >= 1000 DM
        A65 :   unknown/ no savings account

present_emp_since: (qualitative)

    Present employment since
    A71 : unemployed
    A72 :       ... < 1 year
    A73 : 1  <= ... < 4 years  
    A74 : 4  <= ... < 7 years
    A75 :       .. >= 7 years

installment_as_income_perc: (numerical)

    Installment rate in percentage of disposable income

personal_status_sex: (qualitative)

    Personal status and sex
    A91 : male   : divorced/separated
    A92 : female : divorced/separated/married
        A93 : male   : single
    A94 : male   : married/widowed
    A95 : female : single

present_res_since: (numerical)
```
    Present residence since         
```

property: (qualitative)

    Property
    A121 : real estate
    A122 : if not A121 : building society savings agreement/
             life insurance
        A123 : if not A121/A122 : car or other, not in attribute 6
    A124 : unknown / no property

age: (numerical)
```
    Age in years
```

other_installment_plans: (qualitative)

    Other installment plans 
    A141 : bank
    A142 : stores
    A143 : none

housing: (qualitative)

    Housing
    A151 : rent
    A152 : own
    A153 : for free

credits_this_bank : (numerical)

        Number of existing credits at this bank

job : (qualitative)

    Job
    A171 : unemployed/ unskilled  - non-resident
    A172 : unskilled - resident
    A173 : skilled employee / official
    A174 : management/ self-employed/
       highly qualified employee/ officer

people_under_maintenance: (numerical)

    Number of people being liable to provide maintenance for

telephone: (qualitative)

    Telephone
    A191 : none
    A192 : yes, registered under the customers name

foreign_worker: (qualitative)

    foreign worker
    A201 : yes
    A202 : no

class color:
   PURPLE = '\033[95m'
   CYAN = '\033[96m'
   DARKCYAN = '\033[36m'
   BLUE = '\033[94m'
   GREEN = '\033[92m'
   YELLOW = '\033[93m'
   RED = '\033[91m'
   BOLD = '\033[1m'
   UNDERLINE = '\033[4m'
   END = '\033[0m'

print("Size of dataframe is " +color.BOLD+ format(origCreditDf.size) + color.END)
print("Shape(#rows,#columns) of dataframe is "+color.BOLD+ format(origCreditDf.shape) + color.END)
print("Dataframe information \n")
print(origCreditDf.info())

Size of dataframe is 21000
Shape(#rows,#columns) of dataframe is (1000, 21)
Dataframe information 

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 21 columns):
default                       1000 non-null int64
account_check_status          1000 non-null object
duration_in_month             1000 non-null int64
credit_history                1000 non-null object
purpose                       1000 non-null object
credit_amount                 1000 non-null int64
savings                       1000 non-null object
present_emp_since             1000 non-null object
installment_as_income_perc    1000 non-null int64
personal_status_sex           1000 non-null object
other_debtors                 1000 non-null object
present_res_since             1000 non-null int64
property                      1000 non-null object
age                           1000 non-null int64
other_installment_plans       1000 non-null object
housing                       1000 non-null object
credits_this_bank             1000 non-null int64
job                           1000 non-null object
people_under_maintenance      1000 non-null int64
telephone                     1000 non-null object
foreign_worker                1000 non-null object
dtypes: int64(8), object(13)
memory usage: 164.1+ KB
None

#checking for missing values
origCreditDf.isnull().sum()

default                       0
account_check_status          0
duration_in_month             0
credit_history                0
purpose                       0
credit_amount                 0
savings                       0
present_emp_since             0
installment_as_income_perc    0
personal_status_sex           0
other_debtors                 0
present_res_since             0
property                      0
age                           0
other_installment_plans       0
housing                       0
credits_this_bank             0
job                           0
people_under_maintenance      0
telephone                     0
foreign_worker                0
dtype: int64

## Dataset has no missing values. 5 point summary of numerical attributes
origCreditDf.describe().transpose()

obj_origCreditDf=origCreditDf.select_dtypes(include=['object']).copy()
obj_origCreditDf.head(5)
print('defaulters :',origCreditDf['default'].unique())
# Number of 'good' credits (should be 700) and 'bad credits (should be 300)
origCreditDf['default'].value_counts()

defaulters : [0 1]

0    700
1    300
Name: default, dtype: int64

print("Shape(#rows,#columns) of dataframe is "+color.BOLD+ format(obj_origCreditDf.shape) + color.END)
print(obj_origCreditDf.info())
print(obj_origCreditDf.columns)

Shape(#rows,#columns) of dataframe is (1000, 13)
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 13 columns):
account_check_status       1000 non-null object
credit_history             1000 non-null object
purpose                    1000 non-null object
savings                    1000 non-null object
present_emp_since          1000 non-null object
personal_status_sex        1000 non-null object
other_debtors              1000 non-null object
property                   1000 non-null object
other_installment_plans    1000 non-null object
housing                    1000 non-null object
job                        1000 non-null object
telephone                  1000 non-null object
foreign_worker             1000 non-null object
dtypes: object(13)
memory usage: 101.6+ KB
None
Index(['account_check_status', 'credit_history', 'purpose', 'savings',
       'present_emp_since', 'personal_status_sex', 'other_debtors', 'property',
       'other_installment_plans', 'housing', 'job', 'telephone',
       'foreign_worker'],
      dtype='object')

#Let's see possible values of  categrical variables in data

print('account_check_status        :',obj_origCreditDf['account_check_status'].unique())
print('credit_history :',obj_origCreditDf['credit_history'].unique())
print('purpose :',obj_origCreditDf['purpose'].unique())
print('savings :',obj_origCreditDf['savings'].unique())
print('present_emp_since :',obj_origCreditDf['present_emp_since'].unique())
print('personal_status_sex :',obj_origCreditDf['personal_status_sex'].unique())
print('other_debtors :',obj_origCreditDf['other_debtors'].unique())
print('property :',obj_origCreditDf['property'].unique())
print('other_installment_plans :',obj_origCreditDf['other_installment_plans'].unique())
print('housing :',obj_origCreditDf['housing'].unique())
print('job :',obj_origCreditDf['job'].unique())
print('telephone :',obj_origCreditDf['telephone'].unique())
print('foreign_worker :',obj_origCreditDf['foreign_worker'].unique())

account_check_status        : ['< 0 DM' '0 <= ... < 200 DM' 'no checking account'
 '>= 200 DM / salary assignments for at least 1 year']
credit_history : ['critical account/ other credits existing (not at this bank)'
 'existing credits paid back duly till now'
 'delay in paying off in the past'
 'no credits taken/ all credits paid back duly'
 'all credits at this bank paid back duly']
purpose : ['domestic appliances' '(vacation - does not exist?)' 'radio/television'
 'car (new)' 'car (used)' 'business' 'repairs' 'education'
 'furniture/equipment' 'retraining']
savings : ['unknown/ no savings account' '... < 100 DM' '500 <= ... < 1000 DM '
 '.. >= 1000 DM ' '100 <= ... < 500 DM']
present_emp_since : ['.. >= 7 years' '1 <= ... < 4 years' '4 <= ... < 7 years' 'unemployed'
 '... < 1 year ']
personal_status_sex : ['male : single' 'female : divorced/separated/married'
 'male : divorced/separated' 'male : married/widowed']
other_debtors : ['none' 'guarantor' 'co-applicant']
property : ['real estate'
 'if not A121 : building society savings agreement/ life insurance'
 'unknown / no property'
 'if not A121/A122 : car or other, not in attribute 6']
other_installment_plans : ['none' 'bank' 'stores']
housing : ['own' 'for free' 'rent']
job : ['skilled employee / official' 'unskilled - resident'
 'management/ self-employed/ highly qualified employee/ officer'
 'unemployed/ unskilled - non-resident']
telephone : ['yes, registered under the customers name ' 'none']
foreign_worker : ['yes' 'no']

#Lets check correlation among columns of dataframe.
from dython.nominal import associations
corr_df=associations(origCreditDf, nominal_columns=['default','account_check_status', 'credit_history', 'purpose', 'savings',
       'present_emp_since', 'personal_status_sex', 'other_debtors', 'property',
       'other_installment_plans', 'housing', 'job', 'telephone',
       'foreign_worker'], mark_columns=True, theil_u=True, plot=True, return_results=True)

corr_df

# We will ignore very weak correlations
#0.00-0.19: very weak
#0.20-0.39: weak
#0.40-0.59: moderate 
#0.60-0.79: strong
#0.80-1.00: very strong.

corr_triu = corr_df.where(~np.tril(np.ones(corr_df.shape)).astype(np.bool))
corr_triu = corr_triu.stack()
corr_triu[corr_triu > 0.19]

default (nom)              duration_in_month (con)           0.214927
duration_in_month (con)    credit_history (nom)              0.194654
                           purpose (nom)                     0.273692
                           credit_amount (con)               0.624984
                           property (nom)                    0.304274
                           housing (nom)                     0.192174
                           job (nom)                         0.218688
credit_history (nom)       credit_amount (con)               0.193283
                           credits_this_bank (con)           0.595094
purpose (nom)              credit_amount (con)               0.370954
credit_amount (con)        property (nom)                    0.318339
                           housing (nom)                     0.201812
                           job (nom)                         0.334607
                           telephone (nom)                   0.276995
present_emp_since (nom)    present_res_since (con)           0.325431
                           age (con)                         0.409607
personal_status_sex (nom)  age (con)                         0.245809
                           people_under_maintenance (con)    0.284250
present_res_since (con)    property (nom)                    0.191575
                           age (con)                         0.266419
                           housing (nom)                     0.307190
property (nom)             age (con)                         0.224743
age (con)                  housing (nom)                     0.307002
dtype: float64

origCreditDf.columns

Index(['default', 'account_check_status', 'duration_in_month',
       'credit_history', 'purpose', 'credit_amount', 'savings',
       'present_emp_since', 'installment_as_income_perc',
       'personal_status_sex', 'other_debtors', 'present_res_since', 'property',
       'age', 'other_installment_plans', 'housing', 'credits_this_bank', 'job',
       'people_under_maintenance', 'telephone', 'foreign_worker'],
      dtype='object')

#Let's drop irrelevant columns (columns which were part of very weak corelations)
selColumns_CreditDf=origCreditDf.copy()
selColumns_CreditDf=selColumns_CreditDf.drop(["account_check_status","duration_in_month","savings","installment_as_income_perc","other_debtors",
              "other_installment_plans","telephone","foreign_worker"],axis=1)
print(selColumns_CreditDf.info())
print(selColumns_CreditDf.columns)

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 13 columns):
default                     1000 non-null int64
credit_history              1000 non-null object
purpose                     1000 non-null object
credit_amount               1000 non-null int64
present_emp_since           1000 non-null object
personal_status_sex         1000 non-null object
present_res_since           1000 non-null int64
property                    1000 non-null object
age                         1000 non-null int64
housing                     1000 non-null object
credits_this_bank           1000 non-null int64
job                         1000 non-null object
people_under_maintenance    1000 non-null int64
dtypes: int64(6), object(7)
memory usage: 101.6+ KB
None
Index(['default', 'credit_history', 'purpose', 'credit_amount',
       'present_emp_since', 'personal_status_sex', 'present_res_since',
       'property', 'age', 'housing', 'credits_this_bank', 'job',
       'people_under_maintenance'],
      dtype='object')

#we will look into all the boxplot individually to trace out outliers
ax = sns.boxplot(data=selColumns_CreditDf, orient="h")

# Boxplots show presence of outliers as whiskers can be seen. We will treat outlier by using Inter quantile range.
# Let's normalize colmns for age and credit amount using boxcox
from scipy import stats
selColumns_CreditDf['age']= stats.boxcox(selColumns_CreditDf['age'])[0].astype(int)
selColumns_CreditDf['credit_amount']=stats.boxcox(selColumns_CreditDf['credit_amount'])[0].astype(int)
ax = sns.boxplot(data=selColumns_CreditDf, orient="h")

#encoding the categorical variables
encoded_creditdf=pd.get_dummies(selColumns_CreditDf, columns=['credit_history','purpose',
       'present_emp_since','personal_status_sex','property','credits_this_bank', 'housing','job'])
print(encoded_creditdf.info())

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 44 columns):
default                                                                       1000 non-null int64
credit_amount                                                                 1000 non-null int32
present_res_since                                                             1000 non-null int64
age                                                                           1000 non-null int32
people_under_maintenance                                                      1000 non-null int64
credit_history_all credits at this bank paid back duly                        1000 non-null uint8
credit_history_critical account/ other credits existing (not at this bank)    1000 non-null uint8
credit_history_delay in paying off in the past                                1000 non-null uint8
credit_history_existing credits paid back duly till now                       1000 non-null uint8
credit_history_no credits taken/ all credits paid back duly                   1000 non-null uint8
purpose_(vacation - does not exist?)                                          1000 non-null uint8
purpose_business                                                              1000 non-null uint8
purpose_car (new)                                                             1000 non-null uint8
purpose_car (used)                                                            1000 non-null uint8
purpose_domestic appliances                                                   1000 non-null uint8
purpose_education                                                             1000 non-null uint8
purpose_furniture/equipment                                                   1000 non-null uint8
purpose_radio/television                                                      1000 non-null uint8
purpose_repairs                                                               1000 non-null uint8
purpose_retraining                                                            1000 non-null uint8
present_emp_since_.. >= 7 years                                               1000 non-null uint8
present_emp_since_... < 1 year                                                1000 non-null uint8
present_emp_since_1 <= ... < 4 years                                          1000 non-null uint8
present_emp_since_4 <= ... < 7 years                                          1000 non-null uint8
present_emp_since_unemployed                                                  1000 non-null uint8
personal_status_sex_female : divorced/separated/married                       1000 non-null uint8
personal_status_sex_male : divorced/separated                                 1000 non-null uint8
personal_status_sex_male : married/widowed                                    1000 non-null uint8
personal_status_sex_male : single                                             1000 non-null uint8
property_if not A121 : building society savings agreement/ life insurance     1000 non-null uint8
property_if not A121/A122 : car or other, not in attribute 6                  1000 non-null uint8
property_real estate                                                          1000 non-null uint8
property_unknown / no property                                                1000 non-null uint8
credits_this_bank_1                                                           1000 non-null uint8
credits_this_bank_2                                                           1000 non-null uint8
credits_this_bank_3                                                           1000 non-null uint8
credits_this_bank_4                                                           1000 non-null uint8
housing_for free                                                              1000 non-null uint8
housing_own                                                                   1000 non-null uint8
housing_rent                                                                  1000 non-null uint8
job_management/ self-employed/ highly qualified employee/ officer             1000 non-null uint8
job_skilled employee / official                                               1000 non-null uint8
job_unemployed/ unskilled - non-resident                                      1000 non-null uint8
job_unskilled - resident                                                      1000 non-null uint8
dtypes: int32(2), int64(3), uint8(39)
memory usage: 69.4 KB
None

# Split Train/Test data 70:30 ratio
from sklearn.model_selection import train_test_split
#separating target column
y = encoded_creditdf['default']
#removing target column from features
X = encoded_creditdf.loc[:, encoded_creditdf.columns != 'default']
#70:30 train test division
X_train, X_test, y_train, y_test = train_test_split( encoded_creditdf, y, test_size=0.3, random_state=42,)
print(X_train.shape, X_test.shape, y_train.shape, y_test.shape)

(700, 44) (300, 44) (700,) (300,)

# Randomforest Model without parameter tuning
from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(random_state=28)

from pprint import pprint
# Look at parameters used by our current forest
print('Parameters currently in use:\n')
pprint(rf.get_params())

rf=rf.fit(X_train, y_train)
preds = rf.predict_proba(X_test)[:,1]
y_pred=rf.predict(X_test)

Parameters currently in use:

{'bootstrap': True,
 'class_weight': None,
 'criterion': 'gini',
 'max_depth': None,
 'max_features': 'auto',
 'max_leaf_nodes': None,
 'min_impurity_decrease': 0.0,
 'min_impurity_split': None,
 'min_samples_leaf': 1,
 'min_samples_split': 2,
 'min_weight_fraction_leaf': 0.0,
 'n_estimators': 'warn',
 'n_jobs': None,
 'oob_score': False,
 'random_state': 28,
 'verbose': 0,
 'warm_start': False}

C:\Users\phlegmatic\Anaconda3\lib\site-packages\sklearn\ensemble\forest.py:245: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.
  "10 in version 0.20 to 100 in 0.22.", FutureWarning)

#calculate Confusion Matrix
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score

def calculate_confusion_matrix(y_true, y_pred):
    cm=confusion_matrix(y_true, y_pred)
    print(cm)
    
calculate_confusion_matrix(y_test, y_pred)
print(accuracy_score(y_test, y_pred))

[[209   0]
 [  2  89]]
0.9933333333333333

# View a list of the features and their importance scores
importances = rf.feature_importances_
indices = np.argsort(importances)[::-1][:15]
a = encoded_creditdf.columns[:]
features= a.drop('default',1)
#plot it
plt.figure(1)
plt.title('Feature Importances')
plt.barh(range(len(indices)), importances[indices], color='b', align='center')
plt.yticks(range(len(indices)), features[indices])
plt.xlabel('Relative Importance')

Text(0.5, 0, 'Relative Importance')

As we can see credit amount, credit history delay, age are the important features determined by the model to classify the person profile.

trainResult = rf.score(X_train, y_train)
testResult = rf.score(X_test, y_test)
print("Train Accuracy:",(trainResult*100.0))
print("Test Accuracy:" ,(testResult*100.0))

Train Accuracy: 100.0
Test Accuracy: 99.33333333333333

#Hyper Parameter tuning
from sklearn.model_selection import RandomizedSearchCV
# Number of trees in random forest
n_estimators = [int(x) for x in np.linspace(start = 200, stop = 2000, num = 10)]
# Number of features to consider at every split
max_features = ['auto', 'sqrt']
# Maximum number of levels in tree
max_depth = [int(x) for x in np.linspace(10, 110, num = 11)]
max_depth.append(None)
# Minimum number of samples required to split a node
min_samples_split = [2, 5, 10]
# Minimum number of samples required at each leaf node
min_samples_leaf = [1, 2, 4]
# Method of selecting samples for training each tree
bootstrap = [True, False]
# Create the random grid
random_grid = {'n_estimators': n_estimators,
               'max_features': max_features,
               'max_depth': max_depth,
               'min_samples_split': min_samples_split,
               'min_samples_leaf': min_samples_leaf,
               'bootstrap': bootstrap}
pprint(random_grid)

{'bootstrap': [True, False],
 'max_depth': [10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, None],
 'max_features': ['auto', 'sqrt'],
 'min_samples_leaf': [1, 2, 4],
 'min_samples_split': [2, 5, 10],
 'n_estimators': [200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000]}

# Use the random grid to search for best hyperparameters
# First create the base model to tune
rf = RandomForestClassifier()
# Random search of parameters, using 3 fold cross validation, 
# search across 100 different combinations, and use all available cores
rf_random = RandomizedSearchCV(estimator = rf, param_distributions = random_grid, n_iter = 10, cv = 3, verbose=2, random_state=42, n_jobs = -1)
# Fit the random search model
rf_random.fit(X_train, y_train)

Fitting 3 folds for each of 10 candidates, totalling 30 fits

[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.
[Parallel(n_jobs=-1)]: Done  30 out of  30 | elapsed:   20.6s finished

RandomizedSearchCV(cv=3, error_score='raise-deprecating',
                   estimator=RandomForestClassifier(bootstrap=True,
                                                    class_weight=None,
                                                    criterion='gini',
                                                    max_depth=None,
                                                    max_features='auto',
                                                    max_leaf_nodes=None,
                                                    min_impurity_decrease=0.0,
                                                    min_impurity_split=None,
                                                    min_samples_leaf=1,
                                                    min_samples_split=2,
                                                    min_weight_fraction_leaf=0.0,
                                                    n_estimators='warn',
                                                    n_jobs=None,
                                                    oob_sc...
                   param_distributions={'bootstrap': [True, False],
                                        'max_depth': [10, 20, 30, 40, 50, 60,
                                                      70, 80, 90, 100, 110,
                                                      None],
                                        'max_features': ['auto', 'sqrt'],
                                        'min_samples_leaf': [1, 2, 4],
                                        'min_samples_split': [2, 5, 10],
                                        'n_estimators': [200, 400, 600, 800,
                                                         1000, 1200, 1400, 1600,
                                                         1800, 2000]},
                   pre_dispatch='2*n_jobs', random_state=42, refit=True,
                   return_train_score=False, scoring=None, verbose=2)

#best parameters from fitting the random search:
rf_random.best_params_

{'n_estimators': 200,
 'min_samples_split': 10,
 'min_samples_leaf': 2,
 'max_features': 'sqrt',
 'max_depth': 50,
 'bootstrap': True}

#Evaluate Random Search
#To determine if random search yielded a better model, we compare the base model with the best random search model.
def evaluate(model, test_features, test_labels):
    predictions = model.predict(test_features)
    errors = abs(predictions - test_labels)
    mape = 100 * np.mean(errors / test_labels)
    accuracy = 100 - mape
    print('Model Performance')
    print('Average Error: {:0.4f} degrees.'.format(np.mean(errors)))
    print('Accuracy = {:0.2f}%.'.format(accuracy))
    
    return accuracy
base_model = RandomForestClassifier(n_estimators = 10, random_state = 42)
base_model.fit(X_train, y_train)
base_accuracy = evaluate(base_model, X_test, y_test)

Model Performance
Average Error: 0.0167 degrees.
Accuracy = 94.51%.

best_random = rf_random.best_estimator_
random_accuracy = evaluate(best_random, X_test, y_test)
print('Improvement of {:0.2f}%.'.format( 100 * (random_accuracy - base_accuracy) / base_accuracy))

Model Performance
Average Error: 0.0000 degrees.
Accuracy = 100.00%.
Improvement of 5.81%.

from sklearn.model_selection import GridSearchCV
# Create the parameter grid based on the results of random search using 3 folds
param_grid = {
    'bootstrap': [True],
    'max_depth': [40, 50, 60],
    'max_features': [2, 3],
    'min_samples_leaf': [2,3,4],
    'min_samples_split': [8, 10, 12],
    'n_estimators': [100,200]
}
# Create a based model
rf = RandomForestClassifier()
# Instantiate the grid search model with kfold where k=3
grid_search = GridSearchCV(estimator = rf, param_grid = param_grid, 
                          cv = 3, n_jobs = -1, verbose = 2)

# Fit the grid search to the data
grid_search.fit(X_train, y_train)
grid_search.best_params_

Fitting 3 folds for each of 108 candidates, totalling 324 fits

[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.
[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:    2.3s
[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:   12.1s
[Parallel(n_jobs=-1)]: Done 324 out of 324 | elapsed:   24.9s finished
C:\Users\phlegmatic\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py:813: DeprecationWarning: The default of the `iid` parameter will change from True to False in version 0.22 and will be removed in 0.24. This will change numeric results when test-set sizes are unequal.
  DeprecationWarning)

{'bootstrap': True,
 'max_depth': 50,
 'max_features': 3,
 'min_samples_leaf': 2,
 'min_samples_split': 8,
 'n_estimators': 200}

best_grid = grid_search.best_estimator_
grid_accuracy = evaluate(best_grid, X_test, y_test)
print('Improvement of {:0.2f}%.'.format( 100 * (grid_accuracy - base_accuracy) / base_accuracy))

Model Performance
Average Error: 0.0033 degrees.
Accuracy = 98.90%.
Improvement of 4.65%.

num_folds = 5
seed = 28
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
kfold = KFold(n_splits=num_folds, random_state=seed)

results = cross_val_score(best_grid, X_train, y_train, cv=kfold)
print(results)
print("All column cross_val_score: %.3f%% (%.3f%%)" % (results.mean()*100.0, results.std()*100.0))

[0.99285714 0.98571429 1.         1.         0.99285714]
All column cross_val_score: 99.429% (0.535%)

Findings¶

Because we have normalized key column like age, credit_amount using boxcox method. Also the weak corelations between categorical columns allowed us to drop certain columns and make our model simpler.

Hyper parameter tuning using random search as well as GridSearchCV improved the accuracy of our model further. K-fold validation gives 99% accuracy to our model which gives us lot of confidence to classify the person profile as good credit or bad credit.

Number of variables	12
Number of observations	25000
Total Missing (%)	11.0%
Total size in memory	2.3 MiB
Average record size in memory	96.0 B

Numeric	11
Categorical	0
Boolean	1
Date	0
Text (Unique)	0
Rejected	0
Unsupported	0

Distinct count	1582
Unique (%)	6.3%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Mean	5.0544
Minimum	0
Maximum	72.87
Zeros (%)	65.0%

Minimum	0
5-th percentile	0
Q1	0
Median	0
Q3	8.34
95-th percentile	25.7
Maximum	72.87
Range	72.87
Interquartile range	8.34

Dataset Info¶

Domain:¶

Objective:¶

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Overview

Variables

Correlations

Sample

Attribute Information:¶

Findings¶

Standard deviation	9.2618
Coef of variation	1.8324
Kurtosis	5.8703
Mean	5.0544
MAD	6.8295
Skewness	2.2998
Sum	125780
Variance	85.781
Memory size	195.4 KiB

Value	Count	Frequency (%)
0.0	16262	65.0%
10.5	49	0.2%
11.98	44	0.2%
10.84	33	0.1%
11.48	33	0.1%
8.21	28	0.1%
5.64	26	0.1%
11.56	25	0.1%
4.43	24	0.1%
12.42	24	0.1%
Other values (1571)	8338	33.4%
(Missing)	114	0.5%

Distinct count	1404
Unique (%)	5.6%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-0.19
5-th percentile	0
Q1	0
Median	0
Q3	3.01
95-th percentile	30.282
Maximum	90.22
Range	90.41
Interquartile range	3.01

Standard deviation	11.165
Coef of variation	2.6525
Kurtosis	14.38
Mean	4.2091
MAD	6.0853
Skewness	3.7139
Sum	104750
Variance	124.65
Memory size	195.4 KiB

Distinct count	2008
Unique (%)	8.0%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-3.15
5-th percentile	0
Q1	0
Median	0
Q3	18.955
95-th percentile	72.31
Maximum	118.65
Range	121.8
Interquartile range	18.955

Standard deviation	25.637
Coef of variation	1.761
Kurtosis	1.7023
Mean	14.558
MAD	20.163
Skewness	1.6867
Sum	362300
Variance	657.25
Memory size	195.4 KiB

Distinct count	1152
Unique (%)	4.6%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-0.12
5-th percentile	0
Q1	0
Median	0
Q3	0.71
95-th percentile	21.032
Maximum	80.68
Range	80.8
Interquartile range	0.71

Standard deviation	9.1972
Coef of variation	2.8435
Kurtosis	16.97
Mean	3.2344
MAD	5.012
Skewness	3.9549
Sum	80491
Variance	84.588
Memory size	195.4 KiB

Distinct count	1272
Unique (%)	5.1%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Standard deviation	8.2997
Coef of variation	2.3892
Kurtosis	12.986
Mean	3.4738
MAD	5.0468
Skewness	3.3951
Sum	86449
Variance	68.886
Memory size	195.4 KiB

Distinct count	1645
Unique (%)	6.6%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Standard deviation	10.598
Coef of variation	1.7299
Kurtosis	6.5009
Mean	6.1263
MAD	8.0958
Skewness	2.2379
Sum	152460
Variance	112.32
Memory size	195.4 KiB

Distinct count	659
Unique (%)	2.6%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-0.02
5-th percentile	0
Q1	0
Median	0
Q3	0.1
95-th percentile	4.72
Maximum	42.3
Range	42.32
Interquartile range	0.1

Standard deviation	2.7
Coef of variation	3.2628
Kurtosis	51.951
Mean	0.82752
MAD	1.2977
Skewness	6.1467
Sum	20594
Variance	7.2901
Memory size	195.4 KiB

Distinct count	799
Unique (%)	3.2%
Missing (%)	60.5%
Missing (n)	15126
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-3.01
5-th percentile	0.7165
Q1	3.5
Median	4.8
Q3	5.76
95-th percentile	7.49
Maximum	25
Range	28.01
Interquartile range	2.26

Standard deviation	2.2671
Coef of variation	0.48828
Kurtosis	7.2024
Mean	4.6431
MAD	1.5911
Skewness	1.093
Sum	45846
Variance	5.1398
Memory size	195.4 KiB

Distinct count	1045
Unique (%)	4.2%
Missing (%)	67.6%
Missing (n)	16907
Infinite (%)	0.0%
Infinite (n)	0

Minimum	0
5-th percentile	2.212
Q1	5.46
Median	7.29
Q3	8.92
95-th percentile	17.044
Maximum	29.3
Range	29.3
Interquartile range	3.46

Standard deviation	4.1486
Coef of variation	0.53423
Kurtosis	2.7326
Mean	7.7654
MAD	2.8679
Skewness	1.3526
Sum	62846
Variance	17.211
Memory size	195.4 KiB

Distinct count	992
Unique (%)	4.0%
Missing (%)	0.5%
Missing (n)	114
Infinite (%)	0.0%
Infinite (n)	0

Minimum	-68.21
5-th percentile	0
Q1	0
Median	0
Q3	0.33
95-th percentile	8.0075
Maximum	100
Range	168.21
Interquartile range	0.33

Standard deviation	6.8852
Coef of variation	4.1308
Kurtosis	88.788
Mean	1.6668
MAD	2.7103
Skewness	8.1244
Sum	41479
Variance	47.405
Memory size	195.4 KiB

Minimum	26000
5-th percentile	37367
Q1	83022
Median	139880
Q3	196760
95-th percentile	242390
Maximum	253763
Range	227763
Interquartile range	113740

Standard deviation	65731
Coef of variation	0.46992
Kurtosis	-1.199
Mean	139880
MAD	56921
Skewness	6.0424e-05
Sum	3496973366
Variance	4320600000
Memory size	195.4 KiB

	maturity_bond	tag	duration_bond
0	NaN	67922	NaN
1	NaN	134783	NaN
2	NaN	61271	NaN
3	NaN	64412	NaN
4	NaN	184058	NaN